feat: run qdrant and ollama inside thothii

deploy/compose.embedding-gpu.yaml

@@ -0,0 +1,7 @@
+services:
+  embedding:
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - capabilities: ["gpu"]

deploy/env/local.env.example

@@ -13,7 +13,7 @@ THT_WORKSPACE_GIT_AUTHOR_EMAIL=thoth-workspace-registry@example.invalid
 
 THT_DB_NAME=warehouse
 THT_DWH_REST_URL=https://dwh.example.invalid
-THT_VEC_REST_URL=https://vector.example.invalid
-THT_VEC_WRITE_REST_URL=https://vector-write.example.invalid
-THT_OLLAMA_URL=https://embeddings.example.invalid
 THT_LLM_URL=https://llm.example.invalid
+
+# Optional explicit GPU override for Linux hosts that expose a Docker-compatible GPU device.
+# THOTH_ENABLE_EMBEDDING_GPU=1

deploy/env/server.env.example

@@ -18,11 +18,11 @@ THT_WORKSPACE_GIT_AUTHOR_EMAIL=thoth-workspace-registry@example.invalid
 
 THT_DB_NAME=warehouse
 THT_DWH_REST_URL=https://dwh.example.invalid
-THT_VEC_REST_URL=https://vector.example.invalid
-THT_VEC_WRITE_REST_URL=https://vector-write.example.invalid
-THT_OLLAMA_URL=https://embeddings.example.invalid
 THT_LLM_URL=https://llm.example.invalid
 
+# Optional explicit GPU override for Linux hosts that expose a Docker-compatible GPU device.
+# THOTH_ENABLE_EMBEDDING_GPU=1
+
 # Public server session storage. Values are endpoints, roles, or protected source-file paths.
 THT_SESSION_DB_HOST=sessions-db.example.invalid
 THT_SESSION_DB_PORT=5432